Overall, the results in Table 2 make it clear that while a number of other variables are significantly related to analyst coverage, firm size is far and away the dominant factor. Thus in addition to worrying about the influence of these other variables, it is also important to think about potential non-linearities in the relationship between log(l -4- ANALYSTS) and log(SIZE). In this spirit, we proceed as follows. We start in Section 3.B by using the simple size-based regression in Model 1 as our baseline method of generating residual analyst coverage. Next, in Section 3.C, we rerun all of our tests separately for each of the size classes (except the very smallest) in Table 1. In this case, we will each month be running a separate cross-sectional analyst regression for: firms in the 20th-40th NYSE/AMEX percentiles; firms in the 40th-60th percentiles, etc. Among other things, this approach allows the relationship between log(l +ANALYSTS) and log(SIZE) to take on a piecewise linear form, hopefully correcting any deficiencies that arise from imposing an overiy simple linear structure on the entire sample.
In addition, in Section 3.D, we also do sensitivities that take into account the potential for analyst coverage to be correlated with some of the other more significant-looking variables considered in Table 2. For example, we use alternative definitions of residual coverage based on both Model 2, which adds the industry dummies, and Model 8, which adds turnover. And we redo all our tests in terms of beta-adjusted returns, just in case the pronounced relationship between beta and analyst coverage might somehow be affecting the results,